This example shows that the metal-semiconductor junction

diode for similar size and doping has a depletion capacitance sig-
nijicantly smaller than that of a pn-junction, which permits higher
operational frequencies of the device.

6.2 RF Diodes
In this section we will review some practical realizations of the diodes that are
most commonly used in RF and MW circuits. As presented in the previous section, a
classical pn-junction diode is not very suitable for high-frequency applications because
of the high junction capacitance. In contrast, diodes formed by a metal-semiconductor
contact possess smaller junction capacitances and consequently reach higher frequency
limits. Today, Schottky diodes find widespread applications in RF detectors, mixers,
attenuators, oscillators, and amplifiers.
After discussing the Schottky diode in Section 6.2.1, we will continue investigat-
ing a number of special RF diodes. In Section 6.2.2 the PIN diode is analyzed and
placed in context with its primary use as a variable resistor and high-frequency switch.
Besides relying on the rectifier property of diodes, we can also exploit the depen-
dence of the junction capacitance on the applied voltage to construct voltage-controlled
tuning circuits, where diodes are used as variable capacitors. An example of such a spe-
cialized diode is the varactor diode, covered in Section 6.2.3.
At the end of this section we will discuss a few more exotic diode configurations,
such as IMPATT, Tunnel, TRAPATT, BARIUTT, and Gunn diodes, which are less fre-
quently used but which are still of interest due to their unique electric properties.

6.2.1 Schottky Diode

Compared with the conventional pn-junction, the Schottky barrier diode has a dif-
ferent reverse-saturation current mechanism, which is determined by the thermionic
emission of the majority carriers across the potential barrier. This current is orders of
magnitude larger than the diffusion-driven minority carriers constituting the reverse-
saturation current of the ideal pn-junction diode. For instance, the Schottky diode has a
294 Chapter 6 Active RF Components

typical reverse-saturation current density on the order of 10-~A/cm~ compared with

10-"A/cm2 of a conventional Si-based pn-junction diode. The schematic diagram of a
cross-sectional view of the Schottky diode with the corresponding circuit elements is
given in Figure 6- 11.

Metal contact
Depletion I

Metal contact \ Pi

\ ~ e t a l contact \ ~ e t a l contact
Figure 6-11 Cross-sectional view of Si Schottky diode.

The metal electrode (tungsten, aluminum, gold, etc.) is in contact with a weakly
doped n-semiconductor layer epitaxially grown on a highly doped n+ substrate. The
dielectric is assumed to be ideal; that is, the conductance is zero. The current-voltage
characteristic is described by the following equation:

where the reverse-saturation current is given by

and R* is the so-called Richardson constant for thermionic emission of the majority
carrier across the potential barrier. A typical value of R* for Si is 100 A/cm2K2.
The corresponding small-signal equivalent circuit model is illustrated in Figure
6-12. In this circuit we note that the junction resistance R j is dependent on the bias cur-
rent, just as is the diode series resistance, which is comprised of epitaxial and substrate
resistances Rs = Repi+ Rsub. The bond wire inductance is fixed and its value is
approximately on the order of Ls = 0.1 nH . As discussed above, the junction capaci-
tance C, is given by (6.40). Because of the resistance R s , the actual junction voltage is
equal to the applied voltage minus the voltage drop over the diode series resistance,
resulting in the modified exponential expression (6.41).

Figure 6-12 Circuit model of typical Schottky diode.

Typical component values for Schottky diodes are R s = 2 . . . 5 R ,

C , = 0.1 . . . 0.2 pF, and R j = 200 . . . 2 kR. Often, the additional IRs term in
(6.41) is neglected for small bias currents below 0.1 mA.However, for certain applica-
tions, the series resistance may form a feedback loop, which means the resistance is
multiplied by a gain factor of potentially large magnitude. For this situation, the ZRs
term has to be taken into account.
In circuit realizations of high-frequency Schottky diodes, the planar configuration
in Figure 6-1 1 gives rise to relatively large parasitic capacitances for very small metal
contacts of typically 10 pm diameter and less. The stray capacitances can be somewhat
minimized through the addition of an isolation ring, as depicted in Figure 6-13.

Metal contact

n-type epitaxial layer P-type

ring

I n+-type substrate

\ ~ e t a lcontact
Figure 6-13 Schottky diode with additional isolation ring suitable for very-high-
frequency applications.

The small signal junction capacitance and junction resistance can be found by
expanding the electric current expression (6.41) around the quiescent or operating
2% Chapter 6 Active RF Components

point VQ. That means the total diode voltage is written as a DC bias VQ and a small
AC signal carrier frequency component vd :
V = VQ+vd (6.43)
The substitution of (6.43) in (6.41) for a negligible IRs term yields

Expanding this equation in a Taylor series about the Q-point and retaining the first two
terms gives

Here the junction resistance RJ(VQ) is identified as

and the junction capacitance is given by (6.40), with VQ replacing V, .

6.2.2 PIN Diode
PIN diodes find applications as high-frequency switches and variable resistors
(attenuators) in the range from 10 k R to less than 1 R for RF signals up to 50 GHz.
They contain an additional layer of an intrinsic (I-layer) or lightly doped semiconduc-
tor sandwiched between highly doped p+ and n+ layers. Depending upon application
and frequency range, the thickness of the middle layer ranges from 1 to 100 pm. In
forward direction, the diode behaves as if it possesses a variable resistance controlled
by the applied current. However, in reverse direction the lightly doped inner layer cre-
ates space charges whose extent reaches the highly doped outer layers. This effect takes
place even for small reverse voltages and remains essentially constant up to high volt-
ages, with the consequence that the diode behaves similar to a dual plate capacitor. For
instance, a Si-based PIN diode with an internal I-layer of 20 p m and a surface area of
200 by 200 p m has a diffusion capacitance on the order of 0.2 pF.
A generic PIN diode and its practical implementation in mesa processing tech-
nology is presented in Figure 6-14. The advantage of the mesa configuration over the
conventional planar construction is a significant reduction in fringing capacitance.
The mathematical representation of the I-V characteristic depends on the level and
direction of current flow. To keep things simple, we will rely to a large extent on discus-
sions already outlined for the pn-junction.
RF Diodes 297

I n+-type substrate
I
6 6
(a)Simplified structure of a PIN diode (b) Fabrication in mesa processing technology
Figure 6-14 PIN diode construction.

In the forward direction and for a weakly doped n-type intrinsic layer the current
through the diode is

where W is the width of the intrinsic layer; 2 p is the excess minority carrier lifetime,
which can be on the order of up to 7 , = 1 ps ;and ND is the doping concentration in
the middle layer of the lightly doped n-semiconductor. The factor 2 in the exponent
takes into account the presence of two junctions. For a pure intrinsic layer ND = ni,
(6.47) leads to the form

The total charge can be calculated from the relation Q = I ' t p . This allows us to find
the diffusion capacitance:

In the reverse direction, the capacitance is dominated by the bias-dependent space

charge length of the I-layer. For small voltages CJis approximately
298 Chapter 6 Active RF Components

where is the dielectric constant of the intrinsic layer.

The dynamic resistance of a PIN diode can be found through a Taylor series
expansion around the Q-point as already discussed for the Schottky diode. The result is

2
where, with reference to (6.47), we have set Ipo = A(qni W ) / ( N D ~ .p )
Based on the PIN diode's resistive behavior under forward bias ("switch on") and
capacitive behavior under reverse bias ("switch off' or isolation) we can proceed to con-
struct simple small signal models. For the PIN diode in series connection, the electric cir-
cuit model is seen in Figure 6-15 terminated with source and load resistances. The
junction resistance and diffusion capacitance, as derived in (6.49) and (6.50), may in prac-
tice model the PIN diode behavior only very approximately. More quantitative informa-
tion is obtained through measurements or sophisticated computational modeling efforts.

L , - - - - -

(a) Forward bias

VG
T-;?!]
- I

L , - - - - -
I
ZL= zo

(b) Reverse bias (isolation)

Figure 6-15 PIN diode in series connection.

The bias point setting required to operate the PIN diode has to be provided
through a DC circuit that must be separated from the RF signal path. The DC isolation
is achieved by a radio frequency coil (RFC), representing a short circuit at DC and an
open circuit at high frequency. Conversely, blocking capacitors ( C B ) represent an
RF Diodes 299

open circuit at DC and a short circuit at RF. Figure 6-16 shows a typical attenuator cir-
cuit where the PIN diode is used either in series or shunt connection.
Although in the following discussion we will use a DC bias, a low-frequency AC
bias can also be employed. In this case the current through the diode consists of two
components such that I = (dQ/dt) + Q/z, . The implication of this is deferred to the
problem section.

PIN Diode
J

(a) Series connection of PIN diode

wi
n +J
RFC

& -
(b) Shunt connection of PIN diode
Figure 6-16 Attenuator circuit with biased PIN diode in series and shunt
configurations.

For positive DC bias voltage, the series connected PIN diode represents a low
resistance to the RF signal. The shunt connected PIN diode, however, creates a short-cir-
cuit condition, permitting only a negligibly small RF signal to appear at the output port.
 300 Chapter 6 Active RF Components

The shunt connection acts like a high attenuation device with high insertion loss. The
situation is reversed for negative bias condition where the series connected PIN diode
behaves like a capacitor with high impedance or high insertion loss, whereas the shunt
connected diode with a high shunt impedance does not affect the RF signal appreciably.
An often used notation is the transducer loss TL conveniently expressed in terms
of the S parameter SO that with (4.52)

The following example computes the transducer loss for a PIN diode in series
configuration.

& , -0
. w
Example 6-5: Computation of transducer loss of a PIN diode
in series configuration for forward and reverse
bias conditions

Find the transducer loss of a forward and reverse biased PIN diode in
series connection (ZG = ZL = ZO= 50 Q). Assume the junction resis-
tance R j under forward bias ranges between 1 and 20 Q . Further-
more, assume that the reverse bias operating conditions result in the
junction capacitance being CJ = 0.1, 9.3, 0.6, 1.3, and 2.5 pF, and
the frequency range of interest extends from 10 MHz to 50 GHz.

Solution: Based on (6.51) and Figure 6-15, the transducer loss is

found with the aid of the voltage divider rule to be

and

Figure 6-17 plots the transducer loss in dB under forward bias

condition for the given range of junction resistances. In contrast,
Figure 6-18 graphs the reverse bias condition where the PIN diode
essentially has a purely capacitive response.
 Junction resistance R, ,R
Figure 6-17 Transducer loss of series connected PIN diode under forward bias
condition. The diode behaves as a resistor.

Frequency
Figure 6-18 Transducer loss of series connected PIN diode under reverse bias
condition. The diode behaves as a capacitor.
 302 Chapter 6 Actlve RF Componentr

6.2.3 Varactor Diode

The PIN diode with its capacitive behavior under reverse bias already suggests
that a variable capacitance versus voltage characteristic can be created by a specific
middle layer doping profile. A varactor diode exactly accomplishes this task by a suit-
able choice of the intrinsic layer thickness Win addition to selecting a particular doping
distribution ND(x) .

m , ,
Example 6-6: Determination of the required doping profile for
a particular capacitance-voltage behavior

Find the appropriate doping concentration profile ND(x) that

ensures that the varactor diode capacitance changes depending on
the applied reverse biasing voltage as C( V,) = Co'/( V, - V,) ,
where Cot = 5 x 10-l2 FV and the cross-sectional diode area
-4 2
is A = 10 cm .

Solution: The extent of the space charge length can be predicted

based on (6.39) to be

which determines the junction capacitance C = eIA/x. In the deri-

vation of the preceding formula we assumed that the doping concen-
tration in the I-layer is much lower than the doping in the adjacent
layers. If the space charge domain is increased by a small increment
dx, the charge is modified to
a Q = qN,(x)Aax
This differential increase in length can be expressed by a corre-
sponding decrease in capacitance. By differentiating the capacitor
formula, we obtain

ax = - E , A ~ c / c ~
Upon substitution of dx into the expression for dQ and noting that
aQ = CaV,, we have
 2 2
aQ = cav, = -qND(x)A ~ , a c / c
This gives us the desired expression for the doping profile:

For the desired capacitance, we find

Naturally we cannot enforce the doping projle to reach injnity

as x approaches the beginning of the I-layel: Nonetheless, by
approximating a hyperbolic function, it is possible to ensure the
desired capacitance-voltage behavio,:

Figure 6-19 presents the simplified electric circuit model of the varactor diode
consisting of a substrate resistance and voltage-dependent capacitance of the form
( V d i f-f v,)-"~.This is the case when the doping profile is constant. Therefore we
have for the capacitance in generic representation:

where V Q is the reverse bias.

One of the main applications of this diode is the frequency tuning of microwave
circuits. This is due to the fact that the cut-off frequency fv of the first-order varactor
model

can be controlled through the reverse bias V Q.

In addition, the varactor diode can be used to generate short pulses as schematically
explained in Figure 6-20. An applied voltage V A across a series connection of resistor
and diode creates a current flow I " . This current is in phase with the voltage over the pos-
itive cycle. During the negative voltage cycle the stored carriers in the middle layer con-
tribute to the continued current flow until all carriers are removed. At this point the current
304 Chapbr 6 Actfve RF Component.

0.08 1
-2.0 -1.8 -1.6 -1.4 -1.2 -1.0 4 . 8 -0.6 -0.4 -0.2 0
Biasing voltage V, ,V
Figure 6-19 Simplified electric circuit model and capacitance behavior of
varactor diode.

drops abruptly to zero. A transformer can now couple out a voltage pulse according to
Faraday's law V,,, = L(dZv/dt) . The pulse width can be approximated based on the
length of the middle layer Wand the saturation drift velocity vdmax of the injected carrier
concentration.

Figure 6-20 Pulse generation with a varactor diode.

RF Diodes 305

6
If we assume W = 10 pm and vdmm = 10 c d s we obtain a transit time that is
equivalent to a pulse width of

6.2.4 IMPATT Diode

IMPATT stands for IMPact Avalanche and Transit Time diode and exploits the
avalanche effect as originally proposed by Read. The principle of this diode construc-
tion, which is very similar to the PIN diode, is depicted in Figure 6-21. The key differ-
ence is the high electric field strength that is generated at the interface between the n+
and p layer resulting in an avalanche of carriers through impact ionization.

Hole

Impact
I -0

(a) Layer structure and electric field profile (b) Impact ionization
Figure 6-21 IMPATT diode behavior.

The additional ionization current Zion that is generated when the applied RF volt-
age V A produces an electric field that exceeds the critical threshold level is seen in Fig-
ure 6-22. The current slowly decreases during the negative voltage cycle as the excess
carriers are removed. The phase shift between this ionization current and the applied
voltage can be tailored so as to reach 90". The total diode current suffers an additional
delay since the excess carriers have to travel through the intrinsic layer to the p+ layer.
The time constant is dependent on the length and drift velocity as given in (6.47).
Choosing the intrinsic layer length appropriately in conjunction with a suitable doping
concentration can create an additional time delay of 90" .
306 Chapter 6 Active RF Components

Figure 6-22 Applied voltage, ionization current, and total current of an

IMPATT diode.

The electric circuit diagram of an IMPATT device shown in Figure 6-23 is more
intricate than the PIN diode and the reactance reveals an inductive behavior below the
diode's resonance frequency f before turning capacitive above the resonance fre-
quency. The total resistance is positive for f e f and becomes negative for f > f o .

Lion

Figure 6-23 Electric circuit representation for the IMPATT diode.

The resonance frequency is determined based on the operating current IQ , dielec-

tric constant, saturation drift velocity vdmax, and the differential change in the ioniza-
RF Dlodes 307

tion coefficient a with respect to the differential change in electric field strength
a' = a a / a E . The resonance frequency is predicted as

The additional circuit parameters are specified as follows

' d max
R = R,+

where RL is the combined resistance of the semiconductor layers, d is the length of the
avalanche region of the p-layer, and W is the total length, as shown in Figure 6-19. The
negative resistance of this diode above the resonance frequency can be understood in
terms of returning electric energy to the RF or MW resonance circuit; which means the
diode operates as an active device. Thus, the circuit attenuation can be substantially
reduced to the point where additional power is transferred to the load impedance.
Unfortunately, the 180-degree phase shift comes with a price: The efficiency of con-
verting DC to RF power at operating frequencies of 5 to 10 GHz is very low, with typi-
cal values in the range of 10 to 15%.

6.2.5 Tunnel Diode

Tunnel diodes are pn-junction diodes that are made of n and p layers with
extremely high doping (concentrations approach 1 0 ' ~ - 1 0~~ r~n - ~
that
) create very nar-
row space charge zones. This can be seen immediately from equations (6.27) and
(6.28). The result is that the electrons and holes exceed the effective state concentra-
tions in the conduction and valence bands. The Fermi level is shifted into the conduc-
tion band W,, of the nt layer and into the valence band W V p of the pf
semiconductor. We notice from Figure 6-24 that the permissible electron states in either
semiconductor layer are only separated through a very narrow potential barrier.
308 Chapter 6 Active RF Components

Figure 6-24 Tunnel diode and its band energy representation.

Based on quantum mechanical considerations, there is a finite probability that

electrons can be exchanged across the narrow gap rather than having to overcome the
potential barrier through an externally supplied voltage. This phenomenon is known as
tunneling. In thermal equilibrium the electron tunneling from the n to p layer is bal-
anced by the opposite tunneling from the p to n layer. No net current flow results.
The peculiar current-voltage response of the tunnel diode is best explained with
reference to the corresponding energy band deformation for four distinct situations, as
shown in Figure 6-25(b)-(e).

I/ i v7diode current

(a) I-Vcurve of tunnel diode. At high positive biasing voltages the corresponding current
of the tunnel diode approaches the current of the conventional pn-junction diode.
Figure 6-25 Current-voltage behavior of the tunnel diode and comparison with
energy band structure.
RF Diodes 309

(b) Negative current flow for VA < 0 (c) No current flow for V', = 0
pn-junction
Excess of
electrons Tunneling

(d) Positive tunneling current, 0 < VA < Vdin (e) Positive current flow for V, > Vdin
Figure 6-25 Current-voltage behavior of the tunnel diode and comparison with
energy band structure. (Continued)

Unlike the equilibrium condition shown in Figure 6-24 and Figure 6-25(c), for a
negative applied voltage VAa higher concentration of electron states is created in the p-
layer, which results in a higher probability to tunnel into the n-layer than vice versa.
The consequence is that even for small negative voltages, a steep increase in current can
be observed [Figure 6-25(b)]. For a small positive voltage the reservoir of free electrons
is shifted to the n-semiconductor and an increase in free electron states is created in the
p-semiconductor. The consequence is a positive current flow [Figure 6-25(d)] in
response to the tunneling of electrons from the n to the p layer. However, if the applied
voltage reaches a critical value V A= Vdiff no overlapping band structures occur [i.e.,
the condition W c , < W V presponsible for the tunneling effect no longer exists, see Fig-
ure 6-25(e)]. The current flow through the tunnel diode approaches a minimum. Above
 Chapter 6 Active RF Components

this critical voltage point Vdiffthe diode behaves again like a conventional pn-junction
diode and current increases exponentially.
The electric circuit of the tunnel diode, Figure 6-26, is very similar to the IMPATT
diode shown in Figure 6-23. Here Rs and L, are resistance of the semiconductor layer
and associated lead inductance. The junction capacitance CT is in shunt with a negative
conductance -g = d I / d V , which is utilized in the negative slope of the I-V curve
shown in Figure 6-25(a).

cT
Figure 6-26 Electric circuit representation of a tunnel diode.

A simplified amplifier circuit involving a tunnel diode is depicted in Figure 6-27.

If we consider the power amplification factor GTas the ratio of the power delivered to
I 2
the load RL to the maximally available power from the source Ps = V,I /(8 R,) ,we
obtain at resonance

where the influence of Rs is neglected. If g is chosen appropriately (i.e.,

g = 1/ R L + 1/ R G ) , the denominator approaches zero and we have the behavior of an
oscillator.

Figure 6-27 Tunnel diode circuit for amplification/oscillation behavior.

 6.2.6 TRAPATT, BARRITT, and Gunn Diodes
For completeness we briefly mention these additional three diode types without
going into any detail of their circuit representation and quantitative electric parameter
derivations.
The TRApped Plasma Avalanche Triggered Transit (TRAPATT) diode can be
considered an enhancement of the IMPA'IT diode in that a higher efficiency (up to
75%) is realized through the use of bandgap traps. Such traps are energy levels that are
situated inside the bandgap and allow the capture of electrons. External circuits ensure
that during the positive cycle a high barrier voltage is generated, resulting in carrier
multiplication of the electron-hole plasma. The consequence is a breakdown in the rec-
tifier properties of the diode during the negative cycle. The operating frequency is
slightly lower than the IMPAm diode. This is due to the fact that the buildup of the
electron-hole plasma is slower than the transit time through the middle layer in an
IMPA'IT diode.
For the BARRier Injection Transit Time (BARRIT) we are essentially dealing
with a transit time diode whose p+np+ doping profile acts like a transistor without base
contact. The space charge domain extends from the cathode through the middle layer
into the anode. The small-signal circuit model consists of a resistor and shunt capacitor
whose values are dependent on the DC current bias. Unlike the IMPATT diode, this RC
circuit can create a negative phase of up to -90 degrees at a relatively low efficiency of
5% and less. The BARRIT diode finds applications in RADAR mixer and detector
circuits.
The Gunn diode is named after its inventor J. B. Gunn, who found in 1963 that in
certain semiconductors (GaAs, InP) a sufficiently high electric field can cause electrons
to scatter into regions where the bandgap separation increases. As a result of this
increase in bandgap energy, the electrons suffer a loss in mobility p, . This phenome-
non is so dramatic that, for instance in GaAs, the drift velocity (vd = nqp,) can drop
from 2 x lo7 c d s to less than lo7 c d s for electric field strengths growing from
5 kV/cm to 7 kV/cm. The negative differential mobility

is again used for oscillator circuits as we will see in later chapters. To exploit the Gunn
effect for RF and MW applications, a special doping profile is needed to ensure that
once the voltage exceeds the required threshold a stable single-carrier space domain is
created.
312 Chapter 6 Active RF Components

6.3 Bipolar-JunctionTransistor
The transistor was invented in 1948 by Bardeen and Brattain at the former AT&T
Bell Laboratories and has over the past 50 years received a long lists of improvements
and refinements. Initially developed as a point-contact, single device, the transistor has
proliferated into a wide host of sophisticated types ranging from the still popular bipo-
lar junction transistors (BJTs) over the modem GaAs field effect transistors (GaAs
FETs) to the most recent high electron mobility transistors (HEMTs). Although tran-
sistors are often arranged in the millions in integrated circuits (ICs) as part of micropro-
cessor, memory, and peripheral chips, in RF and MW applications the single transistor
has retained its importance. Many RF circuits still rely on discrete transistors in low-
noise, linear, and high-power configurations. It is for this reason that we need to investi-
gate both the DC and RF behavior of the transistors in some detail.
The constituents of a bipolar transistor are three alternatively doped semiconduc-
tors, in npn or pnp configuration. As the word bipolar implies, the internal current flow
is due to both minority and majority carriers. In the following we recapitulate some of
the salient characteristics.

6.3.1 Construction
The BJT is one of the most widely used active RF elements due to its low-cost
construction, relatively high operating frequency, low-noise performance, and high-
power handling capacity. The high-power capacity is achieved through a special inter-
digital emitter-base construction as part of a planar structure. Figure 6-28 shows both
the cross-sectional planar construction and the top view of an interdigital emitter-base
connection.
Because of the interleaved construction shown in Figure 6-28(b) the base-emitter
resistance is kept at a minimum while not compromising the gain performance. As we
will see, a low base resistance directly improves the signal-to-noise ratio by reducing
the current density through the base-emitter junction (shot noise) and by reducing the
random thermal motion in the base (thermal noise), see Chapter 7 for more details.
For frequency applications exceeding 1 GHz it is important to reduce the emitter
width to typically less than 1 pn size while increasing the doping to levels of
loz0, . . 10" ~ m to- both
~ reduce base resistance and increase current gain. Unfortu-
nately, it becomes extremely difficult to ensure the tight tolerances, and self-aligning
processes are required. Furthermore, the acceptor and donor doping concentrations
reach quickly the solubility limits of the Si or GaAs semiconductor materials, providing
a physical limitation of the achievable current gain. For these reasons, heterojunction
bipolar transistors (HBTs) are becoming increasingly popular. HBTs achieve high
-

Bipolar-JunctionTransistor 313

(a) Cross-sectional view of a multifinger bipolar junction transistor

Base bondine uad

well

Emitter bonding pad

(b) Top view of a multifinger bipolar junction transistor
Figure 6-28 lnterdigitated structure of high-frequency BJT.

current gains without having to dope the emitter excessively. Due to additional semi-
conductor layers (for instance, GaAIAs-GaAs sandwich structures) an enhanced elec-
tron injection into the base is achieved while the reverse hole injection into the emitter
is suppressed. The result is an extremely high emitter eficiency as defined by the ratio
of electron current into the base to the sum of the same electron current and reverse
emitter hole current. Figure 6-29 shows a cross-sectional view of such a structure.
Besides GaAs, heterojunctions have been accomplished with InP emitter and
InGaAs base interfaces; even additional heterojunction interfaces between the GaInAs
base and InP collector (double heterojunctions) have been fabricated. The material InP
has the advantage of high breakdown voltage, larger bandgap, and higher thermal con-
ductivity compared to GaAs. Operational frequencies exceeding 100 GHz, and a canier
 Figure 6-29 Cross-sectional view of a GaAs heterojunction bipolar transistor
involving a GaAIAs-GaAs interface.

transition time between base and collector of less than 0.5 ps have been achieved.
Unfortunately, InP is a difficult material to handle and the manufacturing process has
not yet matured to a level that allows it to compete with the Si and GaAs technologies.

6.3.2 Functlonallty
In general, there are two types of BJTs: npn and pnp transistors. The difference
between these two types lies in the doping of the semiconductor used to produce base,
emitter, and collector. For an npn-transistor; collector and emitter are made of n-type
semiconductor, while the base is of p-type. For a pnp-transistor, the semiconductor
types are reversed (n-type for base, and p-type for emitter and collector). Usually, the
emitter has the highest and the base has the lowest concentration of doping atoms. The
BIT is a current-controlled device that is best explained by referring to Figure 6-30,
which shows the structure, elechical symbol, and diode model with associated voltage
~ - - ~ ~ - - ~ .
and current convention for the non-structure. We omit the discussion of the pnp-
transistor since it requires only a reversal of voltage polarity and diode directions.
The first letter in the voltage designation always denotes the positive and the sec-
~ ~

ond letter gives the negative voltage reference points. Under normal mode of operation
(i.e., the forward active mode), the emitter-base diode is operated in forward direction
(with V,, 0.7 V ) and the base-collector diode in reverse. Thus the emitter injects
5

electrons into the base, and conversely from the base a hole current reaches the emitter.
If we maintain the: collector emitter voltage to be larger than the so-called satomticm

dBS 1 p m ) and hgnoy 3 . . . -~~~~..

voltage (typically around 0.1 V), and since the base is a very thin (on the order tof
.... oopea p-type
. layer, only a smau amount~.orc emxmns
~~~-~ .~. - - - -
- . ~ - A
recom-
~~

bine with the holes supplied through the base current. The vast majority of electrons
reach the base-collectorjunction and are collected by the applied reverse voltage V B C .
For the reverse active mode, the collector-emitter voltage is negative (typically
VCE< -0.1 V ) and the base-collector diode is forward biased, while the base-emitter
BipoiarJunction Transistor 315

n-type
collector

P-type
base

n-type
emitter

Figure 6-30 npntransistor: (a) structure with electrical charge flow under forward
active mode of operation, (b) transistor symbol with voltage and current directions,
and (c) diode model.

diode is now operated in reverse direction. Unlike the forward active mode, it is now
the electron flow from the collector that bridges the base and reaches the emitter.
Finally, the saturation mode involves the forward biasing of both the base-emit-
ter and base-collector junctions. This mode typically plays an important role when
dealing with switching circuits.
For a common emitter configuration, Figure 6-31(a) depicts a generic biasing
arrangement, where the base current is fixed through an appropriate choice of biasing
resistor RB and voltage source VBB, resulting in a suitable Q-point. The base current
versus base-emitter voltage, Figure 6-3 l(b), follows a typical diode I-V behavior, which
constitutes the input characteristic of the transistor. The base current and base-emitter
voltage at the intersection point between the load line and the transistor input character-
Q and VBE
istic are identified as IB Q . The collector current versus collector-emitter voltage
behavior as part of the transistor output characteristic follows a more complicated pat-
tern since the collector current must be treated as a parametric curve dependent on the
base current (IB1< IB2. . .) as seen in Figure 6-31(c).
The quantitative BJT behavior is analyzed by investigating the three modes of
operation in terms of setting appropriate operating points and formulating the various
current flows. For simplicity, we will neglect the spatial extent of the individual space
charge domains and assume typical representative voltage and current conditions. To
keep track of the different minoritylmajority and doping conditions in the three semi-
conductor layers, Table 6-3 summarizes the parameters and corresponding notation.
316 Chapter 6 Active RF Components

(a) Biasing circuit for npn BJT in common-emitter configuration

(b) Input characteristic of transistor (c) Output characteristic of transistor

Figure 6-31 Biasing and input, output characteristics of an npn BJT.

Table 6-3 BJT parameter nomenclature

Parameter description I Emitter (mtype) I

I I
Base (ptype)
I
Collector (mtype)

Doping level IN", I4 I N:

E 2 E B 2 B C 2 C
Minority carrier concentration pno = ni / N D nPo = ni / N A pno = ni I N D
in thermal equilibrium
E B C
Majority carrier concentration nno P ~ o n"0
in thermal equilibrium

Spatial extent Id~ I d~ I

BlpolarJunctlon Transistor 317

For the following BJT analysis, it is implicitly understood that the concentrations
obey the inequality pfo << n;o p t o .
Forward Active Mode ( VcE > VcE,, = 0.1 V , I, > 0 )
To find the minority charge concentrations, we consider the configuration shown
in Figure 6-32. Here the concentration is plotted as a function of distance across the
three semiconductor layers. For predicting the spatial minority carrier concentrations in
the respective layer, we rely on the so-called short diode (see Appendix F) analysis,
which approximates the exponentials as linear charge concentration gradients.

Forward biased B Reverse biased

, I 1 *x
x = -dE x =0 xLdB x=dBn+dC
Figure 6-32 Minority carrier concentrations in forward active BJT.

The minority charge concentrations in each layer are given as follows:

E E E E VBE/VT
Emitter: p, (-d,) = p,, and p, (0) = pnoe
B B VBE/VT B B VBC/VT
.Base: np(0) = npoe and np(dB) = npOe =0
c C VBC/VT =
Collector: pn (dB) = pnOe

The last two concentrations are zero because the base-collector voltage is negative (for
instance, for typical transistor values of VCE = 2.5 V and VBE = 0.7 V we find
V,, = -1.8 V , which yields exp[VBc/VT] = exp[-1.8/0.026] + 0). Based on
the aforementioned carrier concentrations we can now predict the diffusion current den-
sity of holes J : ~in~the~ emitter:
318 Chapter 6 Active RF Components

For the diffusion current density of electrons in the base layer J : ~ , we similarly obtain

From the preceding two equations, the collector and base currents can be established as

and

where index F denotes forward current, A is the junction cross-sectional area, and
B B
I, = (qDnnpoA)/dB is the saturation current. The emitter current is directly found
by adding (6.60) and (6.61). The forward current gain PF under constant collector
emitter voltage is defined as

To arrive at (6.62) it is assumed that the exponential function in (6.61) is much larger
than 1, allowing us to neglect the factor -1. Moreover, the ratio between collector and
emitter currents, or a,, is expressed as

C & M W
Example 6-7: Computation of the maximum forward current
gain in a bipolar-junction transistor

Find the maximum forward current gain for a silicon-based BJT

with the following parameters: donor concentration in the emitter,
N~ - l ~ l ~ c m - acceptor
~; concentration in the base.
4-
N A = 10'~cm-'; space charge extent in the emitter,
dE = 0.8 pm ; and space charge extent in the base, dB = 1.2 pm .
Blpolar-Junction Transistor 319

Solution: To apply (6.62), we need to determine the diffusion

constants in base and emitter as described by the Einstein relation
(6.15). Substituting this relation into (6.62), we obtain the forward
current gain:

Furthermore, using the expressions for the minority carrier concen-

trations in base and emitter from Table 6-3, we arrive at the final
expression for p, :

As discussed in Section 6.3.3 and in the following chaptel; the

current gain is only approximately constant. In general, it depends
on the transistor operating conditions and temperature behaviol:

Reverse Active Mode ( V m < -0.1 V, I , > 0 )

The minority carrier concentrations are shown in Figure 6-33 with the associated
space charge domains (i.e., the base-emitter diode is reversed biased whereas the base-
collector diode is forward biased).

Reverse biased Forward biased

1 1 , X

x =-dE x=O x = dB x=dB+dC

Figure 6-33 Reverse active mode of BJT.
 - -
-

320 Chapter 6 Active RF Components

The minority charge concentrations in each layer are as follows:

E E E VBE/VT
Emitter: p, (-dE)= 0 and p, (0) = pnoe =O
B B VBE/VT B B V ~ c / V ~
.Base: np(0) = npoe = 0 and np(dB) = npoe
c
Collector: p, (dB) = pnOe
C VBC/~T c C
and Pn ( d +~dc) = Pn0
From the diffusion current density, we can find the reverse emitter current

and the reverse base current

In a similar manner as done for the forward current gain, we define the reverse current
gain PR

and the collector emitter ratio a,

Saturation Mode ( VBE,VBc > V,, I, > 0 )

This mode of operation implies the forward bias of both diodes, so that the diffu-
sion current density in the base is the combination of forward and reverse carrier flows;
that is, with (6.60) and (6.64):

From (6.68) it is possible to find the emitter current by taking into account the forward
base current. This forward base current (6.61) injects holes into the emitter and thus has
to be taken with a negative sign to comply with our positive emitter current direction
convention. Making the exponential expressions in (6.68) compatible with (6.61), we
add and subtract unity and finally obtain
BlpoiarJunctlon Transistor 321

Because the BJT can be treated as a symmetric device, the collector current is express-
ible in a similar manner as the contribution of three currents: the forward collector and
reverse emitter currents, given by the negative of (6.68), and an additional hole diffu-
sion contribution as the result of the reverse base current IRB. The resulting equation is

Finally, the base current IB = - I , - IE is found from the preceding two equations:

Here again, it is important to recall that the internal emitter current flow is denoted
opposite in sign to the customary external circuit convention.

6.3.3 Frequency Response

The transition frequency f (also known as the cut-off frequency) of a micro-
wave BJT is an important figure of merit since it determines the operating frequency at
which the common-emitter, short-circuit current gain hfe decreases to unity. The tran-
sition frequency f is related to the transit time z that is required for carriers to travel
through the emitter-collector structure:

This transition time is generally composed of three delays:

7 = zE+zB+zC (6.73)
where Z~ , zB , and zc are delays in emitter, base, and collector, respectively. The base-
emitter depletion region charging time is given by

where C E , CC are emitter and collector capacitances, and rE is the emitter resistance
obtained by differentiation of the emitter current with respect to base-emitter voltage.
The second delay in (6.73)is the base layer charging time, and its contribution is given as
322 Chapter 6 Actlve RF Components

where the factor q is doping profile dependent and ranges from q = 2 for uniformly
doped base layers up to q = 60 for highly nonuniform layers. Finally, the transition
time zc through the base-collector junction space charge zone w c can be computed as

with vs representing the saturation drift velocity. In the preceding formulas we have
neglected the collector charging time z~~ = rcCc , which is typically very small when
compared with Z E .
As seen in (6.74a), the emitter charging time is inversely proportional to the collec-
tor current, resulting in higher transition frequencies for increasing collector currents.
However, as the current reaches sufficiently high values, the concentration of charges
injected into the base becomes comparable with the doping level of the base, which
causes an increase of the effective base width and, in turn, reduces the transition fre-
quency. Usually, BJT data sheets provide information about the dependence of the tran-
sition frequency on the collector current. For instance, Figure 6-34 shows the transition
frequency as a function of collector current for the wideband npn-transistor BFG403W
measured at V C E= 2 V , f = 2 GHz, and at an ambient temperature of 25°C.

"1 10
Collector current I,, mA
Figure 6-34 Transition frequency as a function of collector current for the 17 GHz
npn wideband transistor BFG403W (courtesy of Philips Semiconductors).
BlpolarJunctlon Transistor 323

Another aspect of the BJT operated at RF and MW frequencies is that at high fre-
quencies the skin effect physically restricts current flow to the outer perimeter of the
emitter (see also Section 1.4). To keep the charging time as low as possible, the emitter
is constructed in a grid pattern of extremely narrow (less than 1 p m ) strips. Unfortu-
nately, the trade-off is a high current density over the small surface area, limiting the
power handling capabilities. Additional ways to increase the cut-off frequency are to
reduce the base transition time constant Z~ by high doping levels and concomitantly
fabricate very short base layers of less than 100 nm. In addition, a small base thickness
has as an advantage a reduction in power loss.

6.3.4 Temperature Behavior

We have seen in this chapter that almost all parameters describing both the static
and dynamic behavior of semiconductor devices are influenced by the junction temper-
ature Ti . As an example of such a dependence, in Figure 6-35 the forward current gain
PF for a given V C Eis plotted as a function of collector current I c for various junction
temperatures Tj . As we can see from this graph, the current gain raises from 40 at
I c = 3.5 rnA and Tj = -50°C to more than 80 at Ti = 50°C.

-
0 1 2 3 4 5 6
Collector current I,, rnA
Figure 6-35 Current gain P, = a,/(l - a), as a function of collector current
for various junction temperatures at a fixed V C E.
324 Chapter 6 Active RF Components

Another example that shows the strong temperature influence is the dependence
of the input characteristic of a transistor described by the base current as a function of
base-emitter voltage, as depicted in Figure 6-36.

0.5 1.O 1.5 2.0

Base-emitter voltage V,,, V
Figure 6-36 Typical base current as a function of base-emitter voltage for various
junction temperatures at a fixed V,,

Again, if we compare the behavior of the transistor at Tj = -50°C and

Tj = 50°C, we notice that at T, = -50°C and a base-emitter voltage of 1.25 V the
transistor is in cut-off state, whereas at Ti = 50°C the BJT already conducts 4 mA
base current. These two examples underscore the importance of temperature consider-
ations in the design of RF circuits. For instance, the design of a cellular phone for
worldwide use must ensure that our circuit preforms according to specifications under
all temperature conditions encountered by the operator. Standard specifications usually
cover the temperature range from -50°C to 80°C .
The junction temperature also plays an important role when dealing with the max-
imum power dissipation. In general, the manufacturer provides a power derating
curve that specifies the temperature T s up to which the transistor can be operated at
the maximum available power Pto,. For junction temperatures T j exceeding this value,
the power has to be reduced to values dictated by the thermal resistance between the
junction and the soldering point (or case) Rthjs according to
Tjmax-Tj = Tjmar-Tj
P = PtOtT (6.75)
jrnax - T~ 'thjs
where Tim,, is the maximum junction temperature. Typical BJT values vary between
150 and 200°C.
For the RF transistor BFG403W the maximum total power P,,, of 16 mW can be
maintained up to T s = 140°C . For higher temperatures T s IT j < Ti,, , the power
must be derated until the maximum junction temperature T j m , of 150°C is reached.
The corresponding slope is 820°K/W. This value implies that if the power dissipation
of the device decreases by 10 mW, the junction temperature can be increase by 8.2"C
up to the maximum junction temperature. Obviously, transistor cases with such a high
slope (or high thermal resistance) are not acceptable for high-power applications and
manufacturers have to develop effective ways to dissipate the thermal energy generated
by the transistor. Usually, this is done by employing heat sinks and using materials with
high thermal conductivity. Instead of the thermal resistance at the soldering point Rfijhj,,
the manufacturer may supply additional information involving heat resistances between
junction-to-case (Rhjc),case-to-sink (R,,,), and sink-to-air (Rthha)interfaces.
To simplify the thermal analysis it is convenient to resort to a thermal equivalent
circuit with the following correspondences:
Thermal power dissipation = electric current
Temperature = electric voltage
A typical thermal circuit in equilibrium is shown in Figure 6-37, where the total electric
power supplied to the device is balanced through a thermal circuit involving thermal
resistances. In particular, we recognize the thermal resistance of junction to soldering
point which is assumed to be equal to Rhjc.Therefore
T j - T, - 1
Rthjc= Rt h.~ s = - Pw
- -
Y ~ ~ B J T

where junction and soldering point temperatures Ti and T , and thermal power Pw
determine the thermal resistance in Kelvin per Watt ( O W ) , and whose value can also
be expressed in terms of the thermal conductivity yth and the surface area ABm of the
BJT. The solder point temperature is affected by the transition between casing and heat
sink. This constitutes a thermal resistance Rthcswith values up to 5 O W . Finally, the
heat sink represents a thermal resistance of

2
where tjhs is a convection coefficient that can vary widely between 10 W/(K.m ) for
still air, 1 0 0 W / ( ~ . r n ~for
) forced air, up to I O O O W / ( K . ~for
~ )water cooling, and
Ahsis the total area of the heat sink.
326 Chapter 6 Active RF Components

R'hjs Ceramic cover

Figure 6-37 Thermal equivalent circuit of BJT.

The following example provides an often encountered design problem.

M
&
W"
Example 6-8: Thermal analysis involving a BJT mounted on a
heat sink

An RF power BJT generates a total power Pw of 15 W at case tem-

perature of 25°C. The maximum junction temperature is 150°C
and the maximum ambient operating temperature is specified by the
user to be T, = 60°C. What is the maximum dissipated power if
the thermal resistances between case-to-sink and sink-to-air is
2"K/W and 10°K/W, respectively.

Solution: With reference to Figure 6-37, we are dealing with

three thermal resistances: RhjS, Rthcs,and Rthha.The junction-to-
soldering resistance can be found based on equation (6.76):

Adding up all resistances gives us a total thermal resistance of

R,,, = Rhjs + Rthca+ Rthhs= 20.333OK/W
The dissipated power Pthfollows from the temperature drop (junc-
tion temperature Tj minus ambient temperature T,) divided by the
total thermal resistance:
 To operate the BJT in thermal equilibrium, we have to reduce the
total electric power Ptot = Pw to the point where it is in balance with
the computed thermal power Ptot = P,. Thus a reduction from
15 W to 4.43 W is required.

While the design engineer cannot influence the junction-to-sol-

dering point heat resistance, it is the choice of casing and heat sink
that typically allows major improvements in thermal performance.

6.3.5 Limiting Values

The total power dissipation capabilities at a particular temperature restrict the
range of safe operation of the BJT. In our discussion we will exclusively focus on the
active mode in the common-emitter configuration and will neglect the switch-mode
behavior whereby the BJT is operated either in saturation or cut-off mode. For a given
maximum BJT power rating, we can either vary the collector-emitter voltage V C E and
plot the allowable collector current I , = Ptot/Vc, (here we assume that base current
is negligibly small compared to the collector current due to high P ) or vary I , and plot
the allowable collector-emitter voltage VCE = P t o t / I C . The result is the maximum
power hyperbola. This does not mean that Ic and V C E can be increased without
bound. In fact, we need to ensure that I , I I,,, and VCEI VCEmax,as depicted in
Figure 6-38. The safe operating area (SOAR) is defined as a set of biasing points
where the transistor can be operated without risk of unrecoverable damage to the
device. The SOAR domain, shown as a shaded region in Figure 6-38, is more restrictive
than a subset bounded by the maximum power hyperbola, since we have to take into
account two more breakdown mechanisms:
1. Breakdown of first kind. Here the collector current density exhibits a nonuni-
form distribution that results in a local temperature increase, which in turn lowers
the resistance of a portion of the collector domain, creating a channel. The conse-
quence is a further increase in current density through this channel until the posi-
tive feedback begins to destroy the crystal structure (avalanche breakdown),
ultimately destroying the transistor itself.
328 Chapter 6 Actlve RF Components

2. Breakdown of second kind. This breakdown mechanism can take place indepen-
dently of the first mechanism and affects primarily power BJTs. Internal overheat-
ing may cause an abrupt increase in the collector current for constant V C E This.
breakdown mechanism usually occurs at the base-collector junction when the
temperature increases to such high values that the intrinsic concentration is equal
to the collector doping concentration. At this point the resistance of the junction is
abruptly reduced, resulting in a dramatic current increase and melting of the
junction.

Figure 6-38 Operating domain of BJT in active mode with breakdown

mechanisms.

It is interesting to point out that the BJT can exceed the SOAR, indeed even the
maximum power hyperbola, for a short time since the temperature response has a much
larger time constant (on the order of microseconds) in comparison with the electric time
constants.
Additional parameters of importance to a design engineer are the maximum volt-
age conditions for open emitter, base and collector conditions; that is, VCBo (collector-
base voltage, open emitter), VCEo (collector-emitter, open base), and VEBO(emitter-
base voltage, open collector). For instance, values for the BFG403W are as follows:
l = 10 V , Voolmm = 4.5 V , and VEBOl = 1.0 V .
max

6.4 RF Field Effect Transistors

Unlike BJTs, field effect transistors (FETs) are monopolar devices, meaning
that only one carrier type, either holes or electrons, contributes to the current flow
through the channel. If hole contributions are involved we speak of p-channel, other-
wise of n-channel FETs. Moreover, the FET is a voltage-controlled device. A variable
RF Field Effect Transistors 329

electric field controls the current flow from source to drain by changing the applied
voltage on the gate electrode.

6.4.1 Construction
Traditionally FETs are classified according to how the gate is connected to the
conducting channel. Specifically, the following four types are used:
1. Metal Insulator Semiconductor FET (MISFET). Here the gate is separated
from the channel through an insulation layer. One of the most widely used types,
the Metal Oxide Semiconductor FET (MOSFET), belongs to this class.
2. Junction FET (JFET). This type relies on a reverse biased pn-junction that iso-
lates the gate from the channel.
3. MEtal Semiconductor FET (MESFET). If the reverse biased pn-junction is
replaced by a Schottky contact, the channel can be controlled just as in the JFET
case.
4. Hetero FET. As the name implies (and unlike the previous three cases, whose
constructions rely on a single semiconductor material such as Si, GaAs, SiGe, or
InP) the hetero structures utilize abrupt transitions between layers of different
semiconductor materials. Examples are GaAlAs to GaAs or GaInAs to GaAlAs
interfaces. The High Electron Mobility Transistor (HEMT) belongs to this
class.
Figure 6-39 provides an overview of the first three types. In all cases the current
flow is directed from the source to drain, with the gate controlling the current flow.
Due to the presence of a large capacitance formed by the gate electrode and the
insulator or reverse biased pn-junction, MISFETs and J E T S have a relatively low cut-
off frequency and are usually operated in low and medium frequency ranges of typi-
cally up to 1 GHz. GaAs MESFETs find applications up to 60-70 GHz, and HEMT can
operate beyond 100 GHz. Since our interest is geared toward RF applications, the
emphasis will be on the last two types.
In addition to the above physical classification, it is customary to electrically clas-
sify FETs according to enhancement and depletion types. This means that the channel
either experiences an increase in carriers (for instance the n-type channel is injected
with electrons) or a depletion in carriers (for instance the n-type channel is depleted of
electrons). In Figure 6-39 (a) the FET is nonconducting, or normally-off, until a suffi-
ciently positive gate voltage sets up a conduction channel. Normally-off FETs can only
be operated in enhancement mode. Alternatively, normally-on FETs can be of both
enhancement and depletion types.
330 Chapter 6 Active RF Components

Source Gate Drain

I D - m e substrate
\ induced

n-ch-me1 I/
-L-
-
(a) Metal insulator semiconductor FET (MISFET)
Source Gate Drain
Insulator ? ? ?

p+ substrate '/
--
L
-
(b) Junction field effect transistor (JFET)
Source Gate Drain
0 Q Q

(c) Metal semiconductor FET (MESFET)

Figure 6-39 Construction of (a) MISFET, (b) JFET, and (c) MESFET.The shaded
areas depict the space charge domains.
RF Field Effect Transistors 331

6.4.2 Functionality
Because of its importance in RF and MW amplifier, mixer, and oscillator circuits,
we focus our analysis on the MESFET, whose physical behavior is in many ways simi-
lar to the J E T . The analysis is based on the geometry shown in Figure 6-40 where the
transistor is operated in depletion mode.

(a) Operation in the linear region. (b) Operation in the saturation region.
Figure 6-40 Functionality of MESFET for different drain-source voltages.

The Schottky contact builds up a channel space charge domain that affects the
current flow from source to drain. The space charge extent ds can be controlled via the
gate voltage in accordance to our discussion in Section 6.1.3, where (6.39) is adjusted
such that V , is replaced by the gate source voltage V G S:

For instance, the barrier voltage V d is approximately 0.9 V for a GaAs-Au interface.
The resistance R between source and drain is predicted by

with the conductivity given by o = qp,ND and W being the gate width. Substituting
(6.78) into (6.79) yields the drain-current equation:
332 Chapter 6 Actlve RF Components

where we have defined the conductance Go = oWd/L . This equation shows that the
drain current depends linearly on the drain source voltage, a fact that is only true for
small VDs .
As the drain-source voltage increases, the space charge domain near the drain
contact increases as well, resulting in a nonuniform distribution of the depletion region
along the channel; see Figure 6-40(b). If we assume that the voltage along the channel
changes from 0 at the source location to VDs at the drain end, then we can compute the
drain current for the nonuniform space charge region. This approach is also known as
the gradual-channel approximation. The approximation rests primarily on the
assumption that the cross-sectional area at a particular location y along the channel is
given by A(y) = {d - ds(y)) W and the electric field E is only y-directed. The chan-
nel current is thus

where the difference between Vd and VGs in the expression for ds(y) has to be aug-
mented by the additional drop in voltage V(y) along the channel; that is, (6.78)
becomes

Substituting (6.82) into (6.81) and carrying out the integration on both sides of the
equation yields

The result is the output characteristic of the MESFET in terms of the drain current as
a function of VDs for a given fixed VGS, or

This equation reduces for small VDs to (6.80).

An interesting phenomenon occurs when the space charge extends over the entire
channel depth d. The drain-source voltage for this situation is called drain saturation
voltage V,, and is given by
RF Field Effect Transistors

or, explicitly,

'Dsat 7
= q N ~ d-2(V, - VGS) = vp - V, + VGS = VGS- V,, (6.86)
where we introduced the so-called pinch-off voltage Vp = q ~ D d 2 / ( 2 ~and )
threshold voltage V, = V, - V, . The associated drain saturation current is found by
inserting (6.86) into (6.84) with the result

The maximum saturation current in (6.87) is obtained when VGS = 0 , which we

define as IDsat(VGs= 0)= IDss.In Figure 6-41 the typical input/output transfer as
well as the output characteristic behavior is shown.

-1
(a) Circuit symbol (b) Transfer characteristic (c) Output characteristic
Figure 6-41 Transfer and output characteristics of an n-channel MESFET

The saturation drain current (6.87) is often approximated by the simple relation

How well (6.88) approximates (6.87) is discussed in the following example.

m m w
Example 6-9: Drain saturation current in a MESFET

A GaAs MESFET has the following parameters: ND = 1016cm",

d=0.75pm, W = lOpm, L=2pm, E , = 12.0,
V, = 0.8 V , and p, = 8500 cm2/(vs). Determine (a) the pinch-
 - --

Chapter 6 Active RF Components

off voltage, (b) the threshold voltage, (c) the maximum saturation
current IDSs ; and plot the drain saturation current based on (6.87)
and (6.88) for VGS ranging from -4 to 0 V.

Solution: The pinch-off voltage for the FET is independent of

the gate-source voltage and is computed as

Knowing Vp and the barrier voltage Vd = 0.8 V , we find the

threshold voltage to be VTo = Vd - V, = -3.44 V . The maximum
saturation drain current is again independent of the applied drain-
source voltage and, based on (6.87), is equal to

where Go = o q N D W d / L = q 2 p , , ~ i ~ =d 8.16
/ ~ 9.
Figure 6-42 shows results for the saturation drain current com-
puted by using the exact formula (6.87) and by using the quadratic
law approximation given by (6.88).

Quadratic law
approximation

Exact formula

-4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0

Gate-source voltage V,,, V
Figure 6-42 Drain current versus V,, computed using the exact and the
approximate equations (6.87) and (6.88).
RF Field Effect Transistors 335

Because of the excellent agreement, the quadratic law approx-

imation (6.88) is more widely used in the literature and data sheets
than the exact equation.

If VDs reaches the saturation voltage VDsatfor a given VGS,the space charges
pinch off the channel. This implies that the drain current saturates. Interestingly, pinch-
off does not imply a zero IDsince there is no charge barrier impeding the flow of carri-
ers. It is the electric field as a result of the applied voltage VDs that "pulls" the elec-
trons across the depletion space charge domain. Any additional increase VDS> VDsat
will result in a shortening of the channel length from the original length L to the new
length L' = L - AL . The result is that (6.87) must be modified to

The change in channel length as a function of VDs is heuristically taken into account
through the so-called channel length modulation parameter h = AL/(L'VDs) . This
is particularly useful when expressing the drain current in the saturation region:
"Dsat = IDsat(' + "DS) (6.90)
where measurements show a slight increase in drain current as VDsis increased.

c & M w
Example 6-10: I-V characteristic of a MESFET

For discrete gate-source voltages VGS = -1, -1.5, -2, and -2.5 V ,
plot the drain current IDof a MESFET as a function of drain-source
voltage VDs in the range from 0 to 5 V. Assume that the device
parameters are the same as in the previous example and that the
channel length modulation parameter h is set to be 0.03 V-I . Com-
pare your results with the case where h = 0 .

Solution: In the analysis of the MESFET behavior we have to be

careful about choosing the appropriate formulas. At very low drain-
source voltages, the drain current can be described by a simple lin-
ear relation (6.80). As the voltage increases, this approximation
336 Chapter 6 Active RF Components

becomes invalid and a more complicated expression for IDhas to be

employed; see (6.84). Further increase in V, ultimately leads to
channel pinch-off, where VDs 2 V,,, = V,, - V,, . In this case the
drain current is equal to the saturation current given by (6.87). Addi-
tional increases in V, beyond the saturation voltage result only in
minor increases of the drain current due to a shortening of the chan-
nel. At this point, ID is linearly dependent on VDs. Substituting
(6.87) into (6.90) for VDs 2 V,,, , we obtain

To provide a smooth transition from normal to saturation region for

nonzero h we multiply (6.84) by (1 + hVDs). Thus, the final
expression for the drain current for VDs 5 V,, is

The results of applying these formulas to predict IDfor zero (dashed

line) as well as nonzero h (solid line) are shown in Figure 6-43.

-0 1 2 3 4 5
Drain-source voltage V,, ,V
Figure 6-43 Drain current as a function of applied drain-source voltage for
different gate-source biasing conditions.
RF Field Effect Transistors 337

The channel length modulation is similar to the Early effect

encountered in a BJT where the collector current in saturation
mode increases slightly for increasing collector emitter voltage as
discussed in Chapter 7.

6.4.3 Frequency Response

The high-frequency MESFET performance is determined by the transit time of
charge carriers traveling between source and drain and the RC time constant of the
device. Here we will focus our attention on the transit time only and defer the time con-
stant computation, which requires knowledge of the channel capacitance, to Chapter 7.
Since electrons in silicon and GaAs have much higher mobility than holes, n-channel
MESFETs are used in RF and MW applications almost exclusively. Furthermore, since
the electron mobility of GaAs is roughly five times higher than that of Si, GaAs MES-
FETs are usually preferred over Si devices.
The transit time z of the electrons traveling through the channel of gate length L
is computed as

where we have assumed a fixed saturation velocity v,, . As an example, the transition
frequency f = 1/(2nz) for a gate length of 1.0 pm and a saturation velocity of
approximately lo7 cm/s is 15 GHz.

6.4.4 Limiting Values

The MESFET must be operated in a domain limited by maximum drain current
IDmaxr
maximum gate-source voltage VGsmax,and maximum drain-source voltage
VDSma,. The maximum power P,, is dictated by the product of VDs and I,, or

which in turn is related to the channel temperature Tc and ambient temperature T ,

and the thermal resistance between channel and soldering point Rfijs, according to

Figure 6-44 clarifies this point. Also shown in this figure are three possible operat-
ing points. Bias point 3 indicates low amplification and possible clipping of the output
current. However, the power consumption is at a minimum. Bias point 2 reveals accept-
338 Chapter 6 Actlw RF Components

Figure 6-44 Typical maximum output characteristics and three operating points
of MESFET.

able amplification at substantially increased power consumption. Finally, bias point 1

shows high amplification at high power consumption and low output current swing.
Choosing appropriate bias points for specific applications will be investigated in-depth
in subsequent chapters.

6.5 High Electron Mobility Transistors

The high electron mobility transistor (HEMT), also known as modulation-
doped field effect transistor (MODFBT), exploits the differences in band gap energy
between dissimilar semiconductor materials such as GaAlAs and GaAs in an effort to
substantially surpass the upper frequency limit of the MESFET while maintaining low
noise performance and high power rating. At present, transit frequencies of 100 GHz
and above have been achieved. The high frequency behavior is due to a separation of
the carrier electrons from their donor sites at the interface between the doped GaAlAs
and undoped GaAs layer (quantum well), where they are confined to a very narrow
(about 10 nm thick) layer in which motion is possible only parallel to the interface.
Here we speak of a two-dimensional electron gas (2DEG) or plasma of very high
mobility, up to 9000 cm2/(v.s). This is a major improvement over GaAs MESFETs
with pn = 4500 cm2/(v-s) . Because of the thin layer, the canier density is often spec-
ified in terms of a surface density, typically on the order of 10'~-10'~~ m - ~ .
To further reduce carrier scattering by impurities it is customary to insert a spacer
layer ranging between 20 and 100 nm of undoped GaAlAs. The layer is grown through
a molecular beam epitaxial process and has to be sufficiently thin so as to allow the gate
voltage V G S to control the electron plasma through electrostatic force mechanism.
Besides single layer heterostructures (GaAlAs on GaAs), multilayer heterostructures
Hlgh Electron Moblllty Transistors 339

involving several 2DEG channels have also been proposed. As can be expected, manu-
facturing an HEMT is significantly more expensive when compared with the relatively
inexpensive GaAs MESFET due to the precisely controlled thin-layer structures, steep
doping gradients, and the use of more difficult to fabricate semiconductor materials.

6.5.1 Construction
The basic heterostructure is shown in Figure 6-45, where a GaAlAs n-doped
semiconductor is followed by an undoped GaAlAs spacer layer of the same material, an
undoped GaAs layer, and a high resistive semi-insulating (s.i.) GaAl substrate.

Source Gate Drain

? Y ?

Figure 6-45 Generic heterostructure of a depletion-mode HEMT.

The 2DEG is formed in the undoped GaAs layer for zero gate bias condition
because the Fenni level is above the conduction band so that electrons accumulate in
this narrow potential well. As discussed later, the electron concentration can be
depleted by applying an increasingly negative gate voltage.
HEMTs are primarily constructed of heterostructures with matching lattice con-
stants to avoid mechanical tensions between layers. Specific examples are the GaAlAs-
GaAs and InGaAs-InP interfaces. Research is also ongoing with mismatched lattices
whereby, for instance, a larger InGaAs lattice is compressed onto a smaller GaAs lat-
tice. Such device configurations are known as pseudomorphic HEMTs, or pHEMTs.

6.5.2 Functionality
The key issue that determines the drain current flow in a HEMT is the narrow
interface between the GaAlAs and the GaAs layers. For simplicity, we neglect the spacer
layer and concentrate our attention at the energy band model shown in Figure 6-46.
A mathematical model similar to (6.21) can be developed by writing down the
one-dimensional Poisson equation in the form
340 Chapter 6 Active RF Components

.............. ....
.... ....

.......' .. . .
-. .-
................................-3,.
................................... , ."
'.., GaAlAs G~AS,.:.'
* ..................._._
-d 0 X

(a) Energy band diagram (b) Close-up view of conduction band

Figure 6-46 Energy band diagram of GaAIAs-GaAs interface for an HEMT.

where ND and EH are the donor concentration and dielectric constant in the GaAlAs
heterostructure. The boundary conditions for the potential are imposed such that
V(x = 0) = 0 and at the metal-semiconductor side V ( x = -d) = - Vb + VG + AWC/q .
Here V , is the barrier voltage, see (6.38); AWc is the energy difference in the conduc-
tion levels between the n-doped GaAlAs and GaAs; and V G is comprised of the gate-
source voltage as well as the channel voltage drop VG = - VGS + V ( y ) . TO find the
potential, (6.94) is integrated twice. At the metal-semiconductor we set

which yields

where we defined the HEMT threshold voltage VTo as VTo = V, - A W c , q - V, .

Here we have used the previously defined pinch-off voltage Vp = q N D d /(2EH).
From the known electric field at the interface, we find the electron drain current

As mentioned previously, the current flow is restricted to a very thin layer so that it is
appropriate to carry out the integration over a surface charge density Qs at x = 0 . The
High Electron Moblllty Transistors 341

result is o = -y,Q/(WLd) = -pnQS/d. For the surface charge density we find with
Gauss's law Qs = E,E(O) . Inserted in (6.97), we obtain

Upon using (6.96), it is seen that the drain current can be found

Pinch-off occurs when the drain-source voltage is equal to or less than the difference of
gate-source and threshold voltages (i.e., VDs I VGS- VTO).If the equality of this con-
dition is substituted in (6.98c), it is seen

The threshold voltage allows us to determine if the HEMT is operated as an

enhancement or depletion type. For the depletion type we require VTo< 0, or
2
Vb - ( A W c / q ) - Vp < 0 . Substituting the pinch-off voltage Vp = qNDd / ( 2 ~ )and
solving for d, this implies

and if d is less than the preceding expression (i.e., VTO> 0 ), we deal with an enhance-
ment HEMT.

C & M W
Example 6-11: Computation of HEMT-related electric charac-
teristics

Determine typical numerical values for a HEMT device such as

pinch-off voltage, threshold voltage, and drain current for
VGS = -1, -0.75, -0.5, -0.25, and 0 V as a function of drain-
342 Chapter 6 Active RF Components

source voltage VDs. Assume the following parameters:

N D = 10l8 cm-', Vb = 0.81 V , E, = 1 2 . 5 ~ , , d = 50 nm,
AWc = 3.5 x lop2' W.S
2
.
W = 10 p m , L = 0.5 p m , and
pn = 8500 cm /(V-s) .

Solution: The pinch-off voltage of a HEMT is evaluated as

2
Vp = q N D d /(2EH) = 1.81 V
Knowing Vp we can find the threshold voltage as
VTo = Vb - AWc/q - Vp = -1.22 V
Using these values the drain current is computed by relying either on
equation (6.98~) for VDS I VGS - VTO or equation (6.99) for
V,, 2 V,, - VTo . The results of these computations are plotted in
Figure 6-47. We notice in this graph that unlike the GaAs MESFET in
Figure 6-43, a channel length modulation is not taken into account. In
practical simulations such a heuristic adjustment can be added.

Drain-source voltage V, ,V
Figure 6-47 Drain current in a GaAs HEMT.

Both GaAs MESFET and HEMT exhibit similar output charac-

teristics and are thus represented by the same electric circuit model.
Summary 343

6.5.3 Frequency Response

The high-frequency performance of the HEMT is determined by the transit time
similar to the MESFET, However, the transit time z is expressed best through the elec-
tron mobility p, and the electric field E of the drain-source voltage according to

We therefore obtain a transit frequency f = 1 / ( 2 n z ) of a proximately 190 GHz for

?
the gate length of 1.0 pm and a mobility of p, = 8000 cm /(V.s) at a typical drain
voltage V D sof 1.5 V.

6.6 Summary
To understand the functionality and limitations of the most widely employed
active RF solid-state devices, we commenced this chapter with a review of the key ele-
ments of semiconductor physics. The concepts of conduction, valence, and Fermi levels
as part of the energy band model are used as the starting point to examine the various
solid-state mechanisms.
We next turned our attention to the pn-junction, where we derived the barrie~
voltage

and the depletion and diffusion capacitances Cd and Cs in the forms

C, = Co and C - zIo
-e VAN,

JW7Gf - V,
Both capacitances are of primary importance when dealing with the frequency response
of a pn-diode whose current is given by the Shockley equation

This equation underscores the nonlinear current-voltage diode characteristics.

Unlike the pn-junction, the Schottky contact involves an n-type semiconductor
and a metal interface. The Schottky barrier potential Vd is now modified and requires
the work function of metal, q V M, semiconductor, q~ , and the conduction band poten-
tial V c , expressed via